Hello guys, I am running my Unraid server for several years and currently using version 7.3.2.

Since I thought about switching to internal boot, I tried to understand why my Cache drive is at nearly 700 GB of 1 TB for the last years and how to lower the numbers. All my media, except of Nextcloud, is located in the array-pool.

First I checked the shares and found some old / not used docker I removed. It endet at around 622 GB.

After checking AI-search what else I can look into, I installed ncdu (NCurses Disk Usage) to have a better overview of the folders, It showed to following:

ncdu /mnt/cache
1.9 TiB /system → /system/docker/installed/btrfs/subvolumes/... *several folders, 5x 356 GiB, ?x 4 GiB and lower*
96 GiB /domains (VM)
95 GiB /appdata (Docker)
12 GiB /isos
x GiB other stuff

Except for /system I think I understood the folders and files behind it. But with /system I have absolutely no idea.

What adds to the confusion is when I dig deeper into the large folders and look for causes.

1. Big system-folder, all five of them:

   1. ~36 GiB /domains = /Windows 11; /Mint ... I had VM with that names a year ago, currently no Mint-VM and just another Windows 11-VM with a totally differnet name

   2. ~50 GiB /appdata = ~98% /Plex-Media-Server/[...]

      1. ~46% /Metadata

      2. ~21% /Plug-in Support/Databases

      3. ~17% /Media/localhost

      4. ~13% /Cache/PhotoTranscoder

   3. ~270 GiB /system = /docker/installed/btrfs/subvolumes/ → another two big folders, separated 50/50 [...]

      1. ~50% /domains = again /Windows 11; /Mint

      2. ~38% /appdata = again primarily PMS

      3. ~12% /system = "docker...btrfs" loop

2. All of the big folders in /system look similar, all have the btrfs-loop with the same old VM-names in /domains sub-folders and PMS in /appdata sub-folders

I can still convince myself that the total of about 2.1 TiB makes sense with RAID 1, but I don't really understand why it's showing me double the capacity (since the RAID 1 is supposed to mirror the data, there should only be 1 TB available).

Some more details of the cache-pool, with btrfs filesystem usage (after manual balance last night):

Overall:
    Device size:		   1.82TiB
    Device allocated:		   1.14TiB
    Device unallocated:		 696.93GiB
    Device missing:		     0.00B
    Device slack:		     0.00B
    Used:			   1.13TiB
    Free (estimated):		 352.62GiB	(min: 352.62GiB)
    Free (statfs, df):		 352.62GiB
    Data ratio:			      2.00
    Metadata ratio:		      2.00
    Global reserve:		 512.00MiB	(used: 0.00B)
    Multiple profiles:		        no

             Data      Metadata System                               
Id Path      RAID1     RAID1    RAID1     Unallocated Total     Slack
-- --------- --------- -------- --------- ----------- --------- -----
 3 /dev/dm-4 577.00GiB  6.00GiB  32.00MiB   348.46GiB 931.50GiB     -
 4 /dev/dm-5 577.00GiB  6.00GiB  32.00MiB   348.46GiB 931.50GiB     -
-- --------- --------- -------- --------- ----------- --------- -----
   Total     577.00GiB  6.00GiB  32.00MiB   696.93GiB   1.82TiB 0.00B
   Used      572.85GiB  5.15GiB 112.00KiB  

The shares are as follows:

appdata = Cache
backup = Cache
data = Cache_hdd → array ... *different cache drive only for caching data before it is added to the array*
domains = Cache
isos = Cache
local = Cache
nextcloud_Backup = Cache_hdd → array
nextcloud_Data = Cache
system = Cache

The AI also suggested that I run a “docker prune” command and perform a “docker reset.” But since I'd probably end up causing more problems than freeing up storage space, I wanted to ask for your advice and suggestions first.

So if you can explain why /system is so bloated, why I'm still finding old VM data in there, and how I can finally remove ‘everything’ that's unused, both I and my wallet would be very grateful to you.

Thanks in advance to everyone who's helping out.

Edited yesterday at 07:55 AM1 day by undone

Please also post your diagnostics

@MowMdown what part of the diagnostics is needed? - I'd like to share as few files as possible, since some personal information is still visible despite anonymization.

You seem to have Docker in Folder mode, you can't use du on that, it'll just recurse over shared mounts.

Try and see Container usage on the docker page. Also switch the page to advanced mode and remove any orphan images.

Edited 13 hours ago13 hr by Kilrah

@Kilrah you are the best, thanks! It looks like one orphan image took up that much space.

Based purely on the loading time during the removal process, I would attribute it to “netdata” (image ID 75e3630d0e1ad - I typed it out because Unraid wasn't responding for about 5 minutes).

Unraid now shows 277 GB in use for the cache & ncdu shows the following:

ncdu /mnt/cache
192 GiB /system
96 GiB /domains
96 GiB /appdata
...

In /system the biggest folder after the btrfs-loop is just 4.4 GiB, pointing to /var and /tmp

Does anyone can explain what cause that huge files with "netdata"? Or is there some way to check what has caused it?

The answer from @Kilrah will be marked as solution, afterwards.

No reason it'd have been that big so I still suspect misleading usage reporting, as mentioned du on the docker filesystem is useless, and one of the drawbacks of using docker folder is that since it's right on the drive understanding space usage is pretty much impossible.